Code

CAGRA-Q search#2206

Code

Merged

rapids-bot[bot] merged 36 commits intorapidsai:branch-24.04rapidsai/raft:branch-24.04from enp1s0:cagra-qenp1s0/raft:cagra-qCopy head branch name to clipboardMar 21, 2024

+6,795-2,798

Commits

Commits on Mar 18, 2024

• Rebase the PR to enable CI

  

  enp1s0

  authored and

  achirkin

  committed
  35f2971

• Fix broken usage of index.dataset() and the like after the merge of the updated version of CAGRA-Q API

  achirkin

  committed
  dab54e0

• Add explicit instantiations for IVF-PQ search kernels used in tests (rapidsai#2212)

  Show description for 22709e2
  

  tfeher

  authored and

  achirkin

  committed
  22709e2

• Fix style errors

  achirkin

  committed
  80c45f1

• Add explicit instantiations for IVF-PQ search kernels used in tests (rapidsai#2212)

  Show description for 49b6b61
  

  tfeher

  authored and

  achirkin

  committed
  49b6b61

• Merge branch 'branch-24.04' into cagra-q

  achirkin

  authored
  2c02881

Commits on Mar 19, 2024

• Add uint32 VPQ test

  

  enp1s0

  authored and

  achirkin

  committed
  a69f66d

• Update VPQ test

  

  enp1s0

  authored and

  achirkin

  committed
  2d35446

• Merge branch 'branch-24.04' into cagra-q

  achirkin

  authored
  711d3d8

• Cleanup the tests a little bit, add a sanity check for the index type

  achirkin

  committed
  678f767

• Update cpp/include/raft/neighbors/detail/cagra/compute_distance.hpp

  Show description for 09db075
  

  achirkin

  and

  tfeher

  authored
  09db075

• Output an error message if multi_kernel and vpq are specified

  enp1s0

  committed
  4eca385

• Use a raft helper for ceiling division

  achirkin

  committed
  d321873

• Move the instance macro to a separate header to reduce the codesize

  achirkin

  committed
  11f9350

• Merge branch 'branch-24.04' into cagra-q

  achirkin

  authored
  3ff9382

• Use TxN_t

  enp1s0

  committed
  38a8bf2

• Fix incorrect addressing using TxN_t

  achirkin

  committed
  cda2cb8

• Merge branch 'branch-24.04' into cagra-q

  achirkin

  authored
  ff7d3b2

Commits on Mar 20, 2024

• Fix typo

  enp1s0

  committed
  103b9c0

• Fix VPQ search params validation

  enp1s0

  committed
  1fb7c36

• Add dim size validation

  enp1s0

  committed
  89aa91e

• Fix VPQ similarity computation for large dim

  enp1s0

  committed
  daf4f08

• Update CAGRA VPQ test

  enp1s0

  committed
  38ab2bd

• Merge branch 'branch-24.04' into cagra-q

  enp1s0

  authored
  15afe26

• Update cpp/include/raft/neighbors/detail/cagra/cagra_search.cuh

  Show description for 5174811
  

  enp1s0

  and

  achirkin

  authored
  5174811

• Remove redundant team-size and dataset-block-dim parameters from the data descriptor

  achirkin

  committed
  16ddb13

• Mark the strided_dataset::view as deleted (pure virtual) to avoid linker errors

  achirkin

  committed
  317c67f

• Fix the instances in the tests as well

  achirkin

  committed
  59033c7

• Fix a bug in VPQ similarity compute

  enp1s0

  committed
  6567186

• Disable implicit template instantiations for vpq tests

  tfeher

  committed
  ecb896c

• cagra-vpq enable instantiation of int64 kernels

  tfeher

  committed
  1308c61

• Correct copyright year

  tfeher

  committed
  6d663ae

• Update query copy from dmem to smem

  enp1s0

  committed
  0e29876

• Merge branch 'cagra-q' of github.com:enp1s0/raft into cagra-q

  enp1s0

  committed
  31b6982

• Fix query mapping type and usage of a macro that is not available on older cuda

  achirkin

  committed
  6ebb99e

Commits on Mar 21, 2024

• Set pq_len=2 as default, do not allow different pq_len for search

  tfeher

  committed
  b2cdb6d